# install the required packages first
require(jsonlite)
## Loading required package: jsonlite
require(httr)
## Loading required package: httr
require(data.table)
## Loading required package: data.table
get_token <- function(username, password, url_site){
post_body = list(username=username,password=password)
post_url_string = paste0(url_site,'/token/')
result = POST(post_url_string, body = post_body)
# error handling (wrong credentials)
if(result$status_code==400){
print('Check your credentials')
return(0)
}
else if (result$status_code==201){
output = content(result)
token = output$key
}
return(token)
}
get_data <- function(start_date='2020-03-20', token, url_site){
post_body = list(start_date=start_date,username=username,password=password)
post_url_string = paste0(url_site,'/dataset/')
header = add_headers(c(Authorization=paste('Token',token,sep=' ')))
result = GET(post_url_string, header, body = post_body)
output = content(result)
data = data.table::rbindlist(output)
data[,event_date:=as.Date(event_date)]
data = data[order(product_content_id,event_date)]
return(data)
}
send_submission <- function(predictions, token, url_site, submit_now=F){
format_check=check_format(predictions)
if(!format_check){
return(FALSE)
}
post_string="list("
for(i in 1:nrow(predictions)){
post_string=sprintf("%s'%s'=%s",post_string,predictions$product_content_id[i],predictions$forecast[i])
if(i<nrow(predictions)){
post_string=sprintf("%s,",post_string)
} else {
post_string=sprintf("%s)",post_string)
}
}
submission = eval(parse(text=post_string))
json_body = jsonlite::toJSON(submission, auto_unbox = TRUE)
submission=list(submission=json_body)
print(submission)
# {"31515569":2.4,"32939029":2.4,"4066298":2.4,"6676673":2.4,"7061886":2.4,"85004":2.4}
if(!submit_now){
print("You did not submit.")
return(FALSE)
}
header = add_headers(c(Authorization=paste('Token',token,sep=' ')))
post_url_string = paste0(url_site,'/submission/')
result = POST(post_url_string, header, body=submission)
if (result$status_code==201){
print("Successfully submitted. Below you can see the details of your submission")
} else {
print("Could not submit. Please check the error message below, contact the assistant if needed.")
}
print(content(result))
}
check_format <- function(predictions){
if(is.data.frame(predictions) | is.data.frame(predictions)){
if(all(c('product_content_id','forecast') %in% names(predictions))){
if(is.numeric(predictions$forecast)){
print("Format OK")
return(TRUE)
} else {
print("forecast information is not numeric")
return(FALSE)
}
} else {
print("Wrong column names. Please provide 'product_content_id' and 'forecast' columns")
return(FALSE)
}
} else {
print("Wrong format. Please provide data.frame or data.table object")
return(FALSE)
}
}
subm_url = 'http://167.172.183.67'
u_name = "Group17"
p_word = "MuH3NvpH8z8G2Z4l"
submit_now = FALSE
username = u_name
password = p_word
token = get_token(username=u_name, password=p_word, url=subm_url)
data = get_data(token=token,url=subm_url)
library("modeest")
## Registered S3 method overwritten by 'rmutil':
## method from
## print.response httr
library("forecast")
## Registered S3 method overwritten by 'quantmod':
## method from
## as.zoo.data.frame zoo
## Registered S3 method overwritten by 'forecast':
## method from
## predict.default statip
##
## Attaching package: 'forecast'
## The following object is masked from 'package:modeest':
##
## naive
library("ggfortify")
## Loading required package: ggplot2
## Registered S3 methods overwritten by 'ggfortify':
## method from
## autoplot.Arima forecast
## autoplot.acf forecast
## autoplot.ar forecast
## autoplot.bats forecast
## autoplot.decomposed.ts forecast
## autoplot.ets forecast
## autoplot.forecast forecast
## autoplot.stl forecast
## autoplot.ts forecast
## fitted.ar forecast
## fortify.ts forecast
## residuals.ar forecast
library("ggplot2")
library("GGally")
## Registered S3 method overwritten by 'GGally':
## method from
## +.gg ggplot2
#splitting the data by row
la_roche<-data[product_content_id==85004]
sleepy<-data[product_content_id==4066298]
xiaomi_kulaklik<-data[product_content_id==6676673]
fakir_supurge<-data[product_content_id==7061886]
trend_tayt<-data[product_content_id==31515569]
oralb<-data[product_content_id==32939029]
trend_bikini<-data[product_content_id==5926527]
koton_mont<-data[product_content_id==3904356]
#making the new dataframes ts objects
ts_la_roche<-ts(la_roche$sold_count)
ts_sleepy<-ts(sleepy$sold_count)
ts_xiaomi<-ts(xiaomi_kulaklik$sold_count)
ts_fakir<-ts(fakir_supurge$sold_count)
ts_trend_tayt<-ts(trend_tayt$sold_count)
ts_oralb<-ts(oralb$sold_count)
ts_trend_bikini<-ts(trend_bikini$sold_count)
ts_koton<-ts(koton_mont$sold_count)
##plotting ts objects
autoplot(ts_la_roche)
autoplot(ts_sleepy)
autoplot(ts_xiaomi)
autoplot(ts_fakir)
autoplot(ts_trend_tayt)
autoplot(ts_oralb)
autoplot(ts_trend_bikini)
autoplot(ts_koton)
####Fakir supurge
##Building test and train sets
ts_fakir_train<-window(ts_fakir,start=300,end=412)
ts_fakir_test<-window(ts_fakir,start=413,end=432)
##holt
fakir_holt<-holt(ts_fakir_train,20)
summary(fakir_holt)
##
## Forecast method: Holt's method
##
## Model Information:
## Holt's method
##
## Call:
## holt(y = ts_fakir_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.918
## beta = 1e-04
##
## Initial states:
## l = 22.5606
## b = 0.2232
##
## sigma: 21.6685
##
## AIC AICc BIC
## 1235.266 1235.827 1248.903
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set -0.09897731 21.28149 15.13405 -15.52737 41.128 0.9837569
## ACF1
## Training set 0.008463414
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 37.97499 10.2057424 65.74423 -4.494408 80.44438
## 414 38.19711 0.4996482 75.89457 -19.456182 95.85040
## 415 38.41923 -7.0912933 83.92975 -31.183108 108.02157
## 416 38.64135 -13.5262496 90.80895 -41.142108 118.42481
## 417 38.86347 -19.2041697 96.93111 -49.943321 127.67026
## 418 39.08559 -24.3366834 102.50786 -57.910406 136.08158
## 419 39.30771 -29.0520764 107.66750 -65.239561 143.85498
## 420 39.52983 -33.4350773 112.49474 -72.060366 151.12003
## 421 39.75195 -37.5451179 117.04902 -78.463714 157.96761
## 422 39.97407 -41.4257912 121.37393 -84.516275 164.46442
## 423 40.19619 -45.1101980 125.50258 -90.268672 170.66105
## 424 40.41831 -48.6241743 129.46080 -95.760418 176.59704
## 425 40.64043 -51.9883448 133.26921 -101.023057 182.30392
## 426 40.86255 -55.2194858 136.94459 -106.082244 187.80735
## 427 41.08467 -58.3314609 140.50080 -110.959182 193.12852
## 428 41.30679 -61.3358831 143.94947 -115.671633 198.28522
## 429 41.52891 -64.2425957 147.30042 -120.234650 203.29247
## 430 41.75103 -67.0600280 150.56209 -124.661124 208.16319
## 431 41.97315 -69.7954641 153.74177 -128.962196 212.90850
## 432 42.19527 -72.4552499 156.84579 -133.147570 217.53811
autoplot(fakir_holt)
accuracy(fakir_holt,ts_fakir_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -0.09897731 21.28149 15.13405 -15.527366 41.12800 0.9837569
## Test set 18.76486977 43.01421 26.74107 7.868026 35.82515 1.7382474
## ACF1 Theil's U
## Training set 0.008463414 NA
## Test set 0.728084568 1.087074
checkresiduals(fakir_holt)
##
## Ljung-Box test
##
## data: Residuals from Holt's method
## Q* = 18.236, df = 6, p-value = 0.005668
##
## Model df: 4. Total lags used: 10
##ses
fakir_ses<-ses(ts_fakir_train,20)
summary(fakir_ses)
##
## Forecast method: Simple exponential smoothing
##
## Model Information:
## Simple exponential smoothing
##
## Call:
## ses(y = ts_fakir_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.9191
##
## Initial states:
## l = 28.817
##
## sigma: 21.4632
##
## AIC AICc BIC
## 1231.169 1231.390 1239.352
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 0.0858888 21.27238 15.07392 -14.99559 40.7554 0.9798483
## ACF1
## Training set 0.007697116
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 37.73709 10.2309330 65.24325 -4.329947 79.80412
## 414 37.73709 0.3781179 75.09606 -19.398527 94.87271
## 415 37.73709 -7.3716332 82.84581 -31.250748 106.72493
## 416 37.73709 -13.9726869 89.44687 -41.346189 116.82037
## 417 37.73709 -19.8216204 95.29580 -50.291361 125.76554
## 418 37.73709 -25.1287135 100.60289 -58.407859 133.88204
## 419 37.73709 -30.0214029 105.49558 -65.890582 141.36476
## 420 37.73709 -34.5838438 110.05802 -72.868233 148.34241
## 421 37.73709 -38.8750604 114.34924 -79.431083 154.90526
## 422 37.73709 -42.9383443 118.41252 -85.645339 161.11952
## 423 37.73709 -46.8065662 122.28074 -91.561274 167.03545
## 424 37.73709 -50.5053823 125.97956 -97.218125 172.69230
## 425 37.73709 -54.0552736 129.52945 -102.647215 178.12139
## 426 37.73709 -57.4728996 132.94708 -107.874023 183.34820
## 427 37.73709 -60.7720275 136.24621 -112.919604 188.39378
## 428 37.73709 -63.9641900 139.43837 -117.801595 193.27577
## 429 37.73709 -67.0591621 142.53334 -122.534946 198.00912
## 430 37.73709 -70.0653152 145.53949 -127.132461 202.60664
## 431 37.73709 -72.9898835 148.46406 -131.605202 207.07938
## 432 37.73709 -75.8391698 151.31335 -135.962809 211.43699
autoplot(fakir_ses)+autolayer(fitted(fakir_ses))
accuracy(fakir_ses,ts_fakir_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.0858888 21.27238 15.07392 -14.99559 40.75540 0.9798483
## Test set 21.1129108 43.26295 26.55516 14.23702 33.64059 1.7261627
## ACF1 Theil's U
## Training set 0.007697116 NA
## Test set 0.715884135 1.079397
##naive
fakir_naive<-naive(ts_fakir_train,20)
summary(fakir_naive)
##
## Forecast method: Naive method
##
## Model Information:
## Call: naive(y = ts_fakir_train, h = 20)
##
## Residual sd: 21.5198
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0.08035714 21.42366 15.38393 -14.25105 40.98362 1 -0.06574646
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 38 10.5444748 65.45553 -3.989602 79.98960
## 414 38 -0.8279761 76.82798 -21.382265 97.38227
## 415 38 -9.5543646 85.55436 -34.728125 110.72812
## 416 38 -16.9110504 92.91105 -45.979205 121.97920
## 417 38 -23.3924207 99.39242 -55.891605 131.89161
## 418 38 -29.2520274 105.25203 -64.853100 140.85310
## 419 38 -34.6404918 110.64049 -73.094045 149.09405
## 420 38 -39.6559522 115.65595 -80.764530 156.76453
## 421 38 -44.3665756 120.36658 -87.968807 163.96881
## 422 38 -48.8219940 124.82199 -94.782781 170.78278
## 423 38 -53.0596755 129.05968 -101.263756 177.26376
## 424 38 -57.1087292 133.10873 -107.456249 183.45625
## 425 38 -60.9923039 136.99230 -113.395664 189.39566
## 426 38 -64.7291687 140.72917 -119.110706 195.11071
## 427 38 -68.3347919 144.33479 -124.625030 200.62503
## 428 38 -71.8221008 147.82210 -129.958409 205.95841
## 429 38 -75.2020304 151.20203 -135.127566 211.12757
## 430 38 -78.4839283 154.48393 -140.146795 216.14680
## 431 38 -81.6758598 157.67586 -145.028433 221.02843
## 432 38 -84.7848414 160.78484 -149.783210 225.78321
accuracy(fakir_naive,ts_fakir_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.08035714 21.42366 15.38393 -14.25105 40.98362 1.000000
## Test set 20.85000000 43.13525 26.45000 13.63951 33.59629 1.719327
## ACF1 Theil's U
## Training set -0.06574646 NA
## Test set 0.71588414 1.076731
##mean
fakir_mean<-meanf(ts_fakir_train,20)
summary(fakir_mean)
##
## Forecast method: Mean
##
## Model Information:
## $mu
## [1] 54.28319
##
## $mu.se
## [1] 3.257794
##
## $sd
## [1] 34.63083
##
## $bootstrap
## [1] FALSE
##
## $call
## meanf(y = ts_fakir_train, h = 20)
##
## attr(,"class")
## [1] "meanf"
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set -7.530542e-16 34.47725 28.15647 -62.25836 90.4686 1.830252
## ACF1
## Training set 0.8052821
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 54.28319 9.44154 99.12483 -14.6363 123.2027
## 414 54.28319 9.44154 99.12483 -14.6363 123.2027
## 415 54.28319 9.44154 99.12483 -14.6363 123.2027
## 416 54.28319 9.44154 99.12483 -14.6363 123.2027
## 417 54.28319 9.44154 99.12483 -14.6363 123.2027
## 418 54.28319 9.44154 99.12483 -14.6363 123.2027
## 419 54.28319 9.44154 99.12483 -14.6363 123.2027
## 420 54.28319 9.44154 99.12483 -14.6363 123.2027
## 421 54.28319 9.44154 99.12483 -14.6363 123.2027
## 422 54.28319 9.44154 99.12483 -14.6363 123.2027
## 423 54.28319 9.44154 99.12483 -14.6363 123.2027
## 424 54.28319 9.44154 99.12483 -14.6363 123.2027
## 425 54.28319 9.44154 99.12483 -14.6363 123.2027
## 426 54.28319 9.44154 99.12483 -14.6363 123.2027
## 427 54.28319 9.44154 99.12483 -14.6363 123.2027
## 428 54.28319 9.44154 99.12483 -14.6363 123.2027
## 429 54.28319 9.44154 99.12483 -14.6363 123.2027
## 430 54.28319 9.44154 99.12483 -14.6363 123.2027
## 431 54.28319 9.44154 99.12483 -14.6363 123.2027
## 432 54.28319 9.44154 99.12483 -14.6363 123.2027
accuracy(fakir_mean,ts_fakir_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -7.530542e-16 34.47725 28.15647 -62.25836 90.4686 1.830252
## Test set 4.566814e+00 38.03660 28.26327 -23.36638 50.6316 1.837195
## ACF1 Theil's U
## Training set 0.8052821 NA
## Test set 0.7158841 1.143946
##ets
fakir_ets<-ets(ts_fakir_train)
checkresiduals(fakir_ets)
##
## Ljung-Box test
##
## data: Residuals from ETS(M,A,N)
## Q* = 17.292, df = 6, p-value = 0.008268
##
## Model df: 4. Total lags used: 10
fakir_ets_fc<-forecast(fakir_ets,20)
autoplot(fakir_ets_fc)
summary(fakir_ets)
## ETS(M,A,N)
##
## Call:
## ets(y = ts_fakir_train)
##
## Smoothing parameters:
## alpha = 0.7941
## beta = 1e-04
##
## Initial states:
## l = 21.7497
## b = 4.9079
##
## sigma: 0.4052
##
## AIC AICc BIC
## 1228.164 1228.725 1241.801
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -5.956631 22.22949 15.87115 -34.96477 48.83305 1.031671 0.1194258
accuracy(fakir_ets_fc,ts_fakir_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -5.956631 22.22949 15.87115 -34.96477 48.83305 1.031671
## Test set -30.375447 68.29923 60.47045 -125.28339 148.99613 3.930755
## ACF1 Theil's U
## Training set 0.1194258 NA
## Test set 0.8489647 3.314609
##arima
fakir_arima<-auto.arima(ts_fakir_train)
checkresiduals(fakir_arima)
##
## Ljung-Box test
##
## data: Residuals from ARIMA(1,1,1)
## Q* = 13.782, df = 8, p-value = 0.08763
##
## Model df: 2. Total lags used: 10
fakir_arima_fc<-forecast(fakir_arima,20)
autoplot(fakir_arima_fc)
accuracy(fakir_arima_fc,ts_fakir_test)
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0.325362 20.46018 14.52048 -21.10029 42.63872 0.943873 0.04844925
## Test set 21.141301 43.24128 26.54117 14.34107 33.59578 1.725253 0.71507735
## Theil's U
## Training set NA
## Test set 1.078552
##tbats
fakir_tbats<-tbats(ts_fakir_train)
fakir_tbats_fc<-forecast(fakir_tbats,20)
autoplot(fakir_tbats_fc)
accuracy(fakir_tbats_fc,ts_fakir_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.4454263 21.28353 14.86734 -14.79454 40.51517 0.9664201
## Test set 21.5038192 43.45505 26.71153 15.12541 33.70647 1.7363268
## ACF1 Theil's U
## Training set 0.07074501 NA
## Test set 0.71588414 1.083569
##linear regression
autoplot(ts_fakir)
fakir_supurge<-fakir_supurge[,-"product_content_id"]
train.fakir<-fakir_supurge[300:412,]
test.fakir<-fakir_supurge[413:432,]
test.fakir2<-fakir_supurge[413:414,]
fakir_linear<-lm(sold_count~.,data=train.fakir)
fakir.predict<-predict(fakir_linear,test.fakir)
accuracy(fakir.predict,test.fakir$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 3.172859 17.00531 10.90204 2.794273 16.10993
plot(fakir.predict,type="l",xlab="Days",ylab="Fakir Vacuum")
##linear according to correlation
ggcorr(train.fakir,label=TRUE)
## Warning in ggcorr(train.fakir, label = TRUE): data in column(s) 'event_date' are
## not numeric and were ignored
fakir_linear_corr<-lm(sold_count~basket_count+favored_count+visit_count+
category_visits,train.fakir)
fakir.predict.corr<-predict(fakir_linear_corr,test.fakir)
accuracy(fakir.predict.corr,test.fakir$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 6.747498 18.68313 12.20994 11.41426 17.01206
####Xiaomi Kulaklik
##Building test and train sets
ts_xiaomi_train<-window(ts_xiaomi,start=300,end=412)
ts_xiaomi_test<-window(ts_xiaomi,start=413,end=432)
##holt
xiaomi_holt<-holt(ts_xiaomi_train,20)
summary(xiaomi_holt)
##
## Forecast method: Holt's method
##
## Model Information:
## Holt's method
##
## Call:
## holt(y = ts_xiaomi_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.7548
## beta = 1e-04
##
## Initial states:
## l = 310.305
## b = -0.1538
##
## sigma: 137.8548
##
## AIC AICc BIC
## 1653.444 1654.005 1667.081
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 0.01421797 135.3929 84.31978 -9.701816 26.56268 0.9504645
## ACF1
## Training set -0.009278146
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 295.3343 118.66622 472.0024 25.14379 565.5248
## 414 295.1806 73.82390 516.5374 -43.35528 633.7166
## 415 295.0270 36.58798 553.4660 -100.22137 690.2753
## 416 294.8733 4.03429 585.7124 -149.92659 739.6732
## 417 294.7197 -25.26265 614.7020 -194.65106 784.0904
## 418 294.5660 -52.12512 641.2571 -235.65231 824.7843
## 419 294.4123 -77.07857 665.9033 -273.73399 862.5587
## 420 294.2587 -100.48299 689.0004 -309.44661 897.9640
## 421 294.1050 -122.59773 710.8078 -343.18685 931.3969
## 422 293.9514 -143.61703 731.5198 -375.25176 963.1545
## 423 293.7977 -163.69081 751.2862 -405.87060 993.4660
## 424 293.6441 -182.93763 770.2258 -435.22474 1022.5129
## 425 293.4904 -201.45321 788.4340 -463.46054 1050.4413
## 426 293.3367 -219.31614 805.9896 -490.69818 1077.3717
## 427 293.1831 -236.59186 822.9580 -517.03778 1103.4040
## 428 293.0294 -253.33559 839.3944 -542.56376 1128.6226
## 429 292.8758 -269.59440 855.3459 -567.34812 1153.0997
## 430 292.7221 -285.40882 870.8530 -591.45284 1176.8971
## 431 292.5685 -300.81403 885.9509 -614.93174 1200.0687
## 432 292.4148 -315.84083 900.6704 -637.83190 1222.6615
autoplot(xiaomi_holt)
accuracy(xiaomi_holt,ts_xiaomi_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.01421797 135.3929 84.31978 -9.701816 26.56268 0.9504645
## Test set 45.87545350 128.2069 87.73169 4.793077 22.99076 0.9889241
## ACF1 Theil's U
## Training set -0.009278146 NA
## Test set 0.370646079 1.007564
checkresiduals(xiaomi_holt)
##
## Ljung-Box test
##
## data: Residuals from Holt's method
## Q* = 7.654, df = 6, p-value = 0.2646
##
## Model df: 4. Total lags used: 10
##ses
xiaomi_ses<-ses(ts_xiaomi_train,20)
summary(xiaomi_ses)
##
## Forecast method: Simple exponential smoothing
##
## Model Information:
## Simple exponential smoothing
##
## Call:
## ses(y = ts_xiaomi_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.7547
##
## Initial states:
## l = 312.9025
##
## sigma: 136.5981
##
## AIC AICc BIC
## 1649.429 1649.649 1657.611
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set -0.2036101 135.3839 84.33135 -9.77368 26.57569 0.9505949
## ACF1
## Training set -0.009301759
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 295.5373 120.479853 470.5948 27.81002 563.2647
## 414 295.5373 76.215865 514.8588 -39.88591 630.9606
## 415 295.5373 39.493018 551.5817 -96.04866 687.1233
## 416 295.5373 7.413268 583.6614 -145.11040 736.1851
## 417 295.5373 -21.436265 612.5109 -189.23196 780.3066
## 418 295.5373 -47.870656 638.9453 -229.65987 820.7345
## 419 295.5373 -72.410805 663.4855 -267.19079 858.2655
## 420 295.5373 -95.413582 686.4883 -302.37050 893.4452
## 421 295.5373 -117.136149 708.2108 -335.59230 926.6670
## 422 295.5373 -137.771089 728.8458 -367.15071 958.2254
## 423 295.5373 -157.467053 748.5417 -397.27309 988.3478
## 424 295.5373 -176.341634 767.4163 -426.13927 1017.2140
## 425 295.5373 -194.489752 785.5644 -453.89443 1044.9691
## 426 295.5373 -211.989346 803.0640 -480.65774 1071.7324
## 427 295.5373 -228.905338 819.9800 -506.52852 1097.6032
## 428 295.5373 -245.292493 836.3672 -531.59051 1122.6652
## 429 295.5373 -261.197511 852.2722 -555.91514 1146.9898
## 430 295.5373 -276.660597 867.7353 -579.56389 1170.6386
## 431 295.5373 -291.716664 882.7913 -602.59015 1193.6648
## 432 295.5373 -306.396253 897.4709 -625.04065 1216.1153
autoplot(xiaomi_ses)+autolayer(fitted(xiaomi_ses))
accuracy(xiaomi_ses,ts_xiaomi_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -0.2036101 135.3839 84.33135 -9.773680 26.57569 0.9505949
## Test set 44.2126612 127.7934 87.49627 4.250728 23.02930 0.9862703
## ACF1 Theil's U
## Training set -0.009301759 NA
## Test set 0.372875096 1.005521
##naive
xiaomi_naive<-naive(ts_xiaomi_train,20)
summary(xiaomi_naive)
##
## Forecast method: Naive method
##
## Model Information:
## Call: naive(y = ts_xiaomi_train, h = 20)
##
## Residual sd: 140.9433
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0 140.3127 88.71429 -9.091318 27.68069 1 -0.2562068
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 302 122.182090 481.8179 26.99224 577.0078
## 414 302 47.699072 556.3009 -86.91971 690.9197
## 415 302 -9.453757 613.4538 -174.32742 778.3274
## 416 302 -57.635821 661.6358 -248.01552 852.0155
## 417 302 -100.085071 704.0851 -312.93605 916.9361
## 418 302 -138.462127 742.4621 -371.62869 975.6287
## 419 302 -173.753472 777.7535 -425.60215 1029.6021
## 420 302 -206.601855 810.6019 -475.83941 1079.8394
## 421 302 -237.453731 841.4537 -523.02329 1127.0233
## 422 302 -266.634161 870.6342 -567.65090 1171.6509
## 423 302 -294.388540 898.3885 -610.09756 1214.0976
## 424 302 -320.907514 924.9075 -650.65483 1254.6548
## 425 302 -346.342696 950.3427 -689.55459 1293.5546
## 426 302 -370.817013 974.8170 -726.98483 1330.9848
## 427 302 -394.431773 998.4318 -763.10048 1367.1005
## 428 302 -417.271642 1021.2716 -798.03105 1402.0310
## 429 302 -439.408238 1043.4082 -831.88605 1435.8861
## 430 302 -460.902783 1064.9028 -864.75912 1468.7591
## 431 302 -481.808100 1085.8081 -896.73104 1500.7310
## 432 302 -502.170143 1106.1701 -927.87210 1531.8721
accuracy(xiaomi_naive,ts_xiaomi_test)
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0.00 140.3127 88.71429 -9.091318 27.68069 1.0000000 -0.2562068
## Test set 37.75 125.7038 87.15000 2.156931 23.41455 0.9823671 0.3728751
## Theil's U
## Training set NA
## Test set 0.9982734
##mean
xiaomi_mean<-meanf(ts_xiaomi_train,20)
summary(xiaomi_mean)
##
## Forecast method: Mean
##
## Model Information:
## $mu
## [1] 392.3982
##
## $mu.se
## [1] 18.33721
##
## $sd
## [1] 194.9272
##
## $bootstrap
## [1] FALSE
##
## $call
## meanf(y = ts_xiaomi_train, h = 20)
##
## attr(,"class")
## [1] "meanf"
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 2.315877e-14 194.0628 139.219 -24.36656 44.53376 1.569297
## ACF1
## Training set 0.7390088
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 392.3982 139.9972 644.7993 4.469764 780.3267
## 414 392.3982 139.9972 644.7993 4.469764 780.3267
## 415 392.3982 139.9972 644.7993 4.469764 780.3267
## 416 392.3982 139.9972 644.7993 4.469764 780.3267
## 417 392.3982 139.9972 644.7993 4.469764 780.3267
## 418 392.3982 139.9972 644.7993 4.469764 780.3267
## 419 392.3982 139.9972 644.7993 4.469764 780.3267
## 420 392.3982 139.9972 644.7993 4.469764 780.3267
## 421 392.3982 139.9972 644.7993 4.469764 780.3267
## 422 392.3982 139.9972 644.7993 4.469764 780.3267
## 423 392.3982 139.9972 644.7993 4.469764 780.3267
## 424 392.3982 139.9972 644.7993 4.469764 780.3267
## 425 392.3982 139.9972 644.7993 4.469764 780.3267
## 426 392.3982 139.9972 644.7993 4.469764 780.3267
## 427 392.3982 139.9972 644.7993 4.469764 780.3267
## 428 392.3982 139.9972 644.7993 4.469764 780.3267
## 429 392.3982 139.9972 644.7993 4.469764 780.3267
## 430 392.3982 139.9972 644.7993 4.469764 780.3267
## 431 392.3982 139.9972 644.7993 4.469764 780.3267
## 432 392.3982 139.9972 644.7993 4.469764 780.3267
accuracy(xiaomi_mean,ts_xiaomi_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 2.315877e-14 194.0628 139.2190 -24.36656 44.53376 1.569297
## Test set -5.264823e+01 130.9512 107.2889 -27.13062 36.21218 1.209376
## ACF1 Theil's U
## Training set 0.7390088 NA
## Test set 0.3728751 1.214658
##ets
xiaomi_ets<-ets(ts_xiaomi_train)
checkresiduals(xiaomi_ets)
##
## Ljung-Box test
##
## data: Residuals from ETS(M,N,N)
## Q* = 5.1313, df = 8, p-value = 0.7435
##
## Model df: 2. Total lags used: 10
xiaomi_ets_fc<-forecast(xiaomi_ets,20)
autoplot(xiaomi_ets_fc)
summary(xiaomi_ets)
## ETS(M,N,N)
##
## Call:
## ets(y = ts_xiaomi_train)
##
## Smoothing parameters:
## alpha = 0.6367
##
## Initial states:
## l = 290.0437
##
## sigma: 0.2807
##
## AIC AICc BIC
## 1581.417 1581.637 1589.599
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 0.05933439 136.3989 84.37635 -10.22998 26.66883 0.9511022
## ACF1
## Training set 0.1142942
accuracy(xiaomi_ets_fc,ts_xiaomi_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.05933439 136.3989 84.37635 -10.229983 26.66883 0.9511022
## Test set 45.43735676 128.2222 87.61874 4.647509 22.97530 0.9876508
## ACF1 Theil's U
## Training set 0.1142942 NA
## Test set 0.3728751 1.007253
##arima
xiaomi_arima<-auto.arima(ts_xiaomi_train)
checkresiduals(xiaomi_arima)
##
## Ljung-Box test
##
## data: Residuals from ARIMA(2,0,0) with non-zero mean
## Q* = 4.6486, df = 7, p-value = 0.7028
##
## Model df: 3. Total lags used: 10
xiaomi_arima_fc<-forecast(xiaomi_arima,20)
autoplot(xiaomi_arima_fc)
accuracy(xiaomi_arima_fc,ts_xiaomi_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.9314957 129.0497 83.05174 -12.51952 26.66762 0.936171
## Test set -28.6751203 132.8158 100.49252 -19.91566 32.24042 1.132766
## ACF1 Theil's U
## Training set 0.02338572 NA
## Test set 0.46785742 1.183588
##tbats
xiaomi_tbats<-tbats(ts_xiaomi_train)
xiaomi_tbats_fc<-forecast(xiaomi_tbats,20)
autoplot(xiaomi_tbats_fc)
accuracy(xiaomi_tbats_fc,ts_xiaomi_test)
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 3.31667 135.9378 84.26349 -9.269571 26.51935 0.9498301 0.06279922
## Test set 49.75354 129.8145 88.05035 6.045883 22.78502 0.9925161 0.37287510
## Theil's U
## Training set NA
## Test set 1.014251
##linear regression
autoplot(ts_xiaomi)
xiaomi_kulaklik<-xiaomi_kulaklik[,-"product_content_id"]
train.xiaomi<-xiaomi_kulaklik[300:412,]
test.xiaomi<-xiaomi_kulaklik[413:432,]
xiaomi_linear<-lm(sold_count~.,data=train.xiaomi)
xiaomi.predict<-predict(xiaomi_linear,test.xiaomi)
accuracy(xiaomi.predict,test.xiaomi$sold_count)
## ME RMSE MAE MPE MAPE
## Test set -26.67183 42.09958 33.48992 -8.356194 10.061
plot(xiaomi.predict,type="l",xlab="Days",ylab="Xiaomi Earbuds")
##linear according to correlation
ggcorr(train.xiaomi,label=TRUE)
## Warning in ggcorr(train.xiaomi, label = TRUE): data in column(s) 'event_date'
## are not numeric and were ignored
xiaomi_linear_corr<-lm(sold_count~basket_count+favored_count+visit_count
,train.xiaomi)
xiaomi.predict.corr<-predict(xiaomi_linear_corr,test.xiaomi)
accuracy(xiaomi.predict.corr,test.xiaomi$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 27.29032 55.81373 39.44342 5.109304 10.28937
####Sleepy
##Building test and train sets
ts_sleepy_train<-window(ts_sleepy,start=300,end=412)
ts_sleepy_test<-window(ts_sleepy,start=413,end=432)
##holt
sleepy_holt<-holt(ts_sleepy_train,20)
summary(sleepy_holt)
##
## Forecast method: Holt's method
##
## Model Information:
## Holt's method
##
## Call:
## holt(y = ts_sleepy_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.9985
## beta = 1e-04
##
## Initial states:
## l = 151.5011
## b = 3.5023
##
## sigma: 134.5753
##
## AIC AICc BIC
## 1648.002 1648.563 1661.639
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -3.127618 132.172 71.83294 -15.11699 37.67106 0.9984226 0.1837977
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 196.4345 23.96930 368.8996 -67.32824 460.1972
## 414 199.9015 -43.82940 443.6323 -172.85273 572.6557
## 415 203.3685 -95.07948 501.8164 -253.06828 659.8052
## 416 206.8355 -137.75641 551.4274 -320.17233 733.8433
## 417 210.3025 -174.95315 595.5581 -378.89515 799.5001
## 418 213.7695 -208.25679 635.7957 -431.66399 859.2029
## 419 217.2365 -238.61041 673.0833 -479.92117 914.3941
## 420 220.7035 -266.62859 708.0355 -524.60659 966.0135
## 421 224.1705 -292.73826 741.0792 -566.37321 1014.7141
## 422 227.6375 -317.25030 772.5252 -605.69646 1060.9714
## 423 231.1045 -340.39940 802.6083 -642.93528 1105.1442
## 424 234.5715 -362.36790 831.5108 -678.36852 1147.5114
## 425 238.0385 -383.30081 859.3777 -712.21796 1188.2949
## 426 241.5055 -403.31570 886.3266 -744.66340 1227.6743
## 427 244.9725 -422.50946 912.4544 -775.85304 1265.7980
## 428 248.4395 -440.96306 937.8420 -805.91071 1302.7896
## 429 251.9065 -458.74502 962.5579 -834.94118 1338.7541
## 430 255.3735 -475.91386 986.6608 -863.03399 1373.7809
## 431 258.8405 -492.52013 1010.2010 -890.26639 1407.9473
## 432 262.3075 -508.60775 1033.2227 -916.70561 1441.3205
autoplot(sleepy_holt)
accuracy(sleepy_holt,ts_sleepy_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -3.127618 132.1720 71.83294 -15.11699 37.67106 0.9984226
## Test set -24.120966 191.9864 146.43673 -60.50361 82.76518 2.0353579
## ACF1 Theil's U
## Training set 0.1837977 NA
## Test set 0.6699808 1.907687
checkresiduals(sleepy_holt)
##
## Ljung-Box test
##
## data: Residuals from Holt's method
## Q* = 39.777, df = 6, p-value = 5.039e-07
##
## Model df: 4. Total lags used: 10
##ses
sleepy_ses<-ses(ts_sleepy_train,20)
summary(sleepy_ses)
##
## Forecast method: Simple exponential smoothing
##
## Model Information:
## Simple exponential smoothing
##
## Call:
## ses(y = ts_sleepy_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.9999
##
## Initial states:
## l = 80.9111
##
## sigma: 133.1116
##
## AIC AICc BIC
## 1643.585 1643.806 1651.767
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0.9920141 131.9283 71.31219 -12.30396 36.5862 0.9911845 0.184576
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 192.9975 22.40815 363.5868 -67.89639 453.8914
## 414 192.9975 -48.24019 434.2352 -175.94371 561.9387
## 415 192.9975 -102.45220 488.4472 -258.85382 644.8488
## 416 192.9975 -148.15558 534.1506 -328.75111 714.7461
## 417 192.9975 -188.42133 574.4163 -390.33226 776.3273
## 418 192.9975 -224.82450 610.8195 -446.00611 832.0011
## 419 192.9975 -258.30077 644.2958 -497.20364 883.1986
## 420 192.9975 -289.45978 675.4548 -544.85723 930.8522
## 421 192.9975 -318.72501 704.7200 -589.61454 975.6095
## 422 192.9975 -346.40478 732.3998 -631.94710 1017.9421
## 423 192.9975 -372.73187 758.7269 -672.21091 1058.2059
## 424 192.9975 -397.88711 783.8821 -710.68253 1096.6775
## 425 192.9975 -422.01431 808.0093 -747.58189 1133.5769
## 426 192.9975 -445.23006 831.2251 -783.08733 1169.0823
## 427 192.9975 -467.63048 853.6255 -817.34580 1203.3408
## 428 192.9975 -489.29586 875.2909 -850.48014 1236.4751
## 429 192.9975 -510.29414 896.2891 -882.59424 1268.5892
## 430 192.9975 -530.68339 916.6784 -913.77690 1299.7719
## 431 192.9975 -550.51371 936.5087 -944.10477 1330.0998
## 432 192.9975 -569.82871 955.8237 -973.64451 1359.6395
autoplot(sleepy_ses)+autolayer(fitted(sleepy_ses))
accuracy(sleepy_ses,ts_sleepy_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.9920141 131.9283 71.31219 -12.30396 36.58620 0.9911845
## Test set 12.2525021 178.4789 115.24875 -32.49165 56.24902 1.6018689
## ACF1 Theil's U
## Training set 0.1845760 NA
## Test set 0.6316829 1.435676
##naive
sleepy_naive<-naive(ts_sleepy_train,20)
summary(sleepy_naive)
##
## Forecast method: Naive method
##
## Model Information:
## Call: naive(y = ts_sleepy_train, h = 20)
##
## Residual sd: 133.1053
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 1 132.5135 71.94643 -12.41402 36.91139 1 0.1845285
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 193 23.17706 362.8229 -66.72177 452.7218
## 414 193 -47.16590 433.1659 -174.30205 560.3021
## 415 193 -101.14196 487.1420 -256.85131 642.8513
## 416 193 -146.64588 532.6459 -326.44355 712.4435
## 417 193 -186.73564 572.7356 -387.75554 773.7555
## 418 193 -222.97955 608.9795 -443.18582 829.1858
## 419 193 -256.30926 642.3093 -494.15922 880.1592
## 420 193 -287.33181 673.3318 -541.60411 927.6041
## 421 193 -316.46882 702.4688 -586.16532 972.1653
## 422 193 -344.02729 730.0273 -628.31236 1014.3124
## 423 193 -370.23897 756.2390 -668.39967 1054.3997
## 424 193 -395.28392 781.2839 -706.70261 1092.7026
## 425 193 -419.30532 805.3053 -743.44017 1129.4402
## 426 193 -442.41926 828.4193 -778.78989 1164.7899
## 427 193 -464.72142 850.7214 -812.89810 1198.8981
## 428 193 -486.29176 872.2918 -845.88709 1231.8871
## 429 193 -507.19792 893.1979 -877.86030 1263.8603
## 430 193 -527.49771 913.4977 -908.90616 1294.9062
## 431 193 -547.24103 933.2410 -939.10096 1325.1010
## 432 193 -566.47127 952.4713 -968.51108 1354.5111
accuracy(sleepy_naive,ts_sleepy_test)
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 1.00 132.5135 71.94643 -12.41402 36.91139 1.000000 0.1845285
## Test set 12.25 178.4787 115.25000 -32.49337 56.25039 1.601886 0.6316829
## Theil's U
## Training set NA
## Test set 1.435697
##mean
sleepy_mean<-meanf(ts_sleepy_train,20)
summary(sleepy_mean)
##
## Forecast method: Mean
##
## Model Information:
## $mu
## [1] 223.1504
##
## $mu.se
## [1] 15.45001
##
## $sd
## [1] 164.2359
##
## $bootstrap
## [1] FALSE
##
## $call
## meanf(y = ts_sleepy_train, h = 20)
##
## attr(,"class")
## [1] "meanf"
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -6.536991e-15 163.5075 84.0838 -29.77285 45.65213 1.1687 0.6710025
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 223.1504 10.49003 435.8109 -103.6985 549.9994
## 414 223.1504 10.49003 435.8109 -103.6985 549.9994
## 415 223.1504 10.49003 435.8109 -103.6985 549.9994
## 416 223.1504 10.49003 435.8109 -103.6985 549.9994
## 417 223.1504 10.49003 435.8109 -103.6985 549.9994
## 418 223.1504 10.49003 435.8109 -103.6985 549.9994
## 419 223.1504 10.49003 435.8109 -103.6985 549.9994
## 420 223.1504 10.49003 435.8109 -103.6985 549.9994
## 421 223.1504 10.49003 435.8109 -103.6985 549.9994
## 422 223.1504 10.49003 435.8109 -103.6985 549.9994
## 423 223.1504 10.49003 435.8109 -103.6985 549.9994
## 424 223.1504 10.49003 435.8109 -103.6985 549.9994
## 425 223.1504 10.49003 435.8109 -103.6985 549.9994
## 426 223.1504 10.49003 435.8109 -103.6985 549.9994
## 427 223.1504 10.49003 435.8109 -103.6985 549.9994
## 428 223.1504 10.49003 435.8109 -103.6985 549.9994
## 429 223.1504 10.49003 435.8109 -103.6985 549.9994
## 430 223.1504 10.49003 435.8109 -103.6985 549.9994
## 431 223.1504 10.49003 435.8109 -103.6985 549.9994
## 432 223.1504 10.49003 435.8109 -103.6985 549.9994
accuracy(sleepy_mean,ts_sleepy_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -6.536991e-15 163.5075 84.0838 -29.77285 45.65213 1.16870
## Test set -1.790044e+01 178.9553 130.3252 -53.19147 72.84882 1.81142
## ACF1 Theil's U
## Training set 0.6710025 NA
## Test set 0.6316829 1.716916
##ets
sleepy_ets<-ets(ts_sleepy_train)
checkresiduals(sleepy_ets)
##
## Ljung-Box test
##
## data: Residuals from ETS(M,A,N)
## Q* = 18.773, df = 6, p-value = 0.004564
##
## Model df: 4. Total lags used: 10
sleepy_ets_fc<-forecast(sleepy_ets,20)
autoplot(sleepy_ets_fc)
summary(sleepy_ets)
## ETS(M,A,N)
##
## Call:
## ets(y = ts_sleepy_train)
##
## Smoothing parameters:
## alpha = 0.8978
## beta = 3e-04
##
## Initial states:
## l = 89.3627
## b = 28.4785
##
## sigma: 0.4228
##
## AIC AICc BIC
## 1573.877 1574.438 1587.514
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -30.09652 137.9725 77.615 -31.29353 43.24843 1.078789 0.2457719
accuracy(sleepy_ets_fc,ts_sleepy_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -30.09652 137.9725 77.6150 -31.29353 43.24843 1.078789
## Test set -275.00767 406.2965 368.5761 -253.67636 268.36253 5.122924
## ACF1 Theil's U
## Training set 0.2457719 NA
## Test set 0.8070913 6.225621
##arima
sleepy_arima<-auto.arima(ts_sleepy_train)
checkresiduals(sleepy_arima)
##
## Ljung-Box test
##
## data: Residuals from ARIMA(0,0,2) with non-zero mean
## Q* = 4.5785, df = 7, p-value = 0.7112
##
## Model df: 3. Total lags used: 10
sleepy_arima_fc<-forecast(sleepy_arima,20)
autoplot(sleepy_arima_fc)
accuracy(sleepy_arima_fc,ts_sleepy_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.5166868 110.1224 62.60735 -17.90589 36.53934 0.870194
## Test set -15.2465969 179.3098 131.18022 -51.83645 72.72038 1.823304
## ACF1 Theil's U
## Training set 0.03399085 NA
## Test set 0.65584251 1.703929
sleepy_arima_fc2<-forecast(sleepy_arima,2)
ts_sleepy_test2<-window(ts_sleepy_test,end=414)
accuracy(sleepy_arima_fc,ts_sleepy_test[1:2])
## ME RMSE MAE MPE MAPE MASE
## Training set 0.5166868 110.1224 62.60735 -17.90589 36.53934 0.870194
## Test set 343.5949378 429.5196 343.59494 51.93037 51.93037 4.775705
## ACF1
## Training set 0.03399085
## Test set NA
##tbats
sleepy_tbats<-tbats(ts_sleepy_train)
sleepy_tbats_fc<-forecast(sleepy_tbats,20)
autoplot(sleepy_tbats_fc)
accuracy(sleepy_tbats_fc,ts_sleepy_test)
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 10.82399 118.8437 60.77630 -12.337497 31.77580 0.8447438 0.3717874
## Test set 69.86837 194.9753 95.34392 6.210551 30.02737 1.3252071 0.6277320
## Theil's U
## Training set NA
## Test set 1.198207
##linear regression
autoplot(ts_sleepy)
sleepy<-sleepy[,-"product_content_id"]
train.sleepy<-sleepy[300:412,]
test.sleepy<-sleepy[413:432,]
sleepy_linear<-lm(sold_count~.,data=train.sleepy)
sleepy.predict<-predict(sleepy_linear,test.sleepy)
accuracy(sleepy.predict,test.sleepy$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 0.1126242 46.35828 31.11812 -6.344007 15.87181
plot(sleepy.predict)
##linear according to correlation
ggcorr(train.sleepy,label=TRUE)
## Warning in ggcorr(train.sleepy, label = TRUE): data in column(s) 'event_date'
## are not numeric and were ignored
sleepy_linear_corr<-lm(sold_count~basket_count+favored_count+visit_count+
category_visits,train.sleepy)
sleepy.predict.corr<-predict(sleepy_linear_corr,test.sleepy)
accuracy(sleepy.predict.corr,test.sleepy$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 23.00037 53.35012 32.68619 8.195822 14.97151
plot(sleepy.predict.corr,type="l",xlab="Days",ylab="Sleepy Wet Towel")
####La Roche
##Building test and train sets
ts_la_roche_train<-window(ts_la_roche,start=300,end=412)
ts_la_roche_test<-window(ts_la_roche,start=413,end=432)
##holt
ts_la_roche_train<-window(ts_la_roche,start=300,end=412)
ts_la_roche_test<-window(ts_la_roche,start=413,end=432)
la_roche_holt<-holt(ts_la_roche_train,20)
summary(la_roche_holt)
##
## Forecast method: Holt's method
##
## Model Information:
## Holt's method
##
## Call:
## holt(y = ts_la_roche_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.7254
## beta = 1e-04
##
## Initial states:
## l = 5.4763
## b = 0.2199
##
## sigma: 18.5148
##
## AIC AICc BIC
## 1199.719 1200.280 1213.356
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set -0.01254865 18.18414 13.08172 -14.45196 36.83587 0.993324
## ACF1
## Training set 0.04919043
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 30.06596 6.3383000 53.79362 -6.222365 66.35429
## 414 30.28575 0.9711234 59.60037 -14.547101 75.11860
## 415 30.50553 -3.4911449 64.50221 -21.487900 82.49897
## 416 30.72532 -7.3835088 68.83415 -27.557105 89.00775
## 417 30.94511 -10.8744255 72.76464 -33.012349 94.90256
## 418 31.16489 -14.0628132 76.39260 -38.004915 100.33470
## 419 31.38468 -17.0126273 79.78199 -42.632615 105.40198
## 420 31.60447 -19.7680457 82.97698 -46.963011 110.17195
## 421 31.82425 -22.3610996 86.00961 -51.045093 114.69360
## 422 32.04404 -24.8158906 88.90397 -54.915720 119.00380
## 423 32.26383 -27.1510934 91.67875 -58.603452 123.13111
## 424 32.48361 -29.3815266 94.34875 -62.130954 127.09818
## 425 32.70340 -31.5191831 96.92598 -65.516565 130.92336
## 426 32.92319 -33.5739309 99.42030 -68.775378 134.62175
## 427 33.14297 -35.5540058 101.83995 -71.919990 138.20594
## 428 33.36276 -37.4663660 104.19189 -74.961040 141.68656
## 429 33.58255 -39.3169533 106.48205 -77.907617 145.07271
## 430 33.80233 -41.1108896 108.71556 -80.767554 148.37222
## 431 34.02212 -42.8526270 110.89687 -83.547659 151.59190
## 432 34.24191 -44.5460643 113.02988 -86.253896 154.73771
autoplot(la_roche_holt)
accuracy(la_roche_holt,ts_la_roche_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -0.01254865 18.18414 13.08172 -14.451960 36.83587 0.993324
## Test set 10.84606646 24.10274 15.91868 8.553503 31.54323 1.208740
## ACF1 Theil's U
## Training set 0.04919043 NA
## Test set 0.53323940 0.9744738
checkresiduals(la_roche_holt)
##
## Ljung-Box test
##
## data: Residuals from Holt's method
## Q* = 16.321, df = 6, p-value = 0.01213
##
## Model df: 4. Total lags used: 10
##ses
la_roche_ses<-ses(ts_la_roche_train,20)
summary(la_roche_ses)
##
## Forecast method: Simple exponential smoothing
##
## Model Information:
## Simple exponential smoothing
##
## Call:
## ses(y = ts_la_roche_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.7279
##
## Initial states:
## l = 10.8691
##
## sigma: 18.3405
##
## AIC AICc BIC
## 1195.636 1195.856 1203.818
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 0.2289621 18.17745 13.04502 -13.97565 36.26303 0.9905371
## ACF1
## Training set 0.04766302
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 29.70209 6.1978253 53.20636 -6.244582 65.64877
## 414 29.70209 0.6303517 58.77383 -14.759298 74.16348
## 415 29.70209 -4.0304088 63.43459 -21.887316 81.29150
## 416 29.70209 -8.1211435 67.52533 -28.143555 87.54774
## 417 29.70209 -11.8107093 71.21489 -33.786259 93.19044
## 418 29.70209 -15.1981107 74.60229 -38.966842 98.37103
## 419 29.70209 -18.3472965 77.75148 -43.783106 103.18729
## 420 29.70209 -21.3024104 80.70659 -48.302563 107.70675
## 421 29.70209 -24.0954432 83.49963 -52.574138 111.97832
## 422 29.70209 -26.7504573 86.15464 -56.634632 116.03882
## 423 29.70209 -29.2860920 88.69028 -60.512550 119.91673
## 424 29.70209 -31.7171345 91.12132 -64.230509 123.63469
## 425 29.70209 -34.0555500 93.45973 -67.806807 127.21099
## 426 29.70209 -36.3111827 95.71537 -71.256499 130.66068
## 427 29.70209 -38.4922477 97.89643 -74.592150 133.99633
## 428 29.70209 -40.6056849 100.00987 -77.824374 137.22856
## 429 29.70209 -42.6574203 102.06160 -80.962232 140.36642
## 430 29.70209 -44.6525619 104.05675 -84.013538 143.41772
## 431 29.70209 -46.5955494 105.99973 -86.985081 146.38926
## 432 29.70209 -48.4902709 107.89445 -89.882807 149.28699
autoplot(la_roche_ses)+autolayer(fitted(la_roche_ses))
accuracy(la_roche_ses,ts_la_roche_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.2289621 18.17745 13.04502 -13.97565 36.26303 0.9905371
## Test set 13.2979082 24.62386 16.21916 16.29590 30.41988 1.2315568
## ACF1 Theil's U
## Training set 0.04766302 NA
## Test set 0.49813737 0.9763132
##naive
la_roche_naive<-naive(ts_la_roche_train,20)
summary(la_roche_naive)
##
## Forecast method: Naive method
##
## Model Information:
## Call: naive(y = ts_la_roche_train, h = 20)
##
## Residual sd: 18.7636
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0.09821429 18.67987 13.16964 -13.36197 36.88527 1 -0.184464
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 23 -0.9392152 46.93922 -13.61187 59.61187
## 414 23 -10.8551628 56.85516 -28.77700 74.77700
## 415 23 -18.4639370 64.46394 -40.41362 86.41362
## 416 23 -24.8784303 70.87843 -50.22374 96.22374
## 417 23 -30.5297124 76.52971 -58.86663 104.86663
## 418 23 -35.6388620 81.63886 -66.68040 112.68040
## 419 23 -40.3372099 86.33721 -73.86590 119.86590
## 420 23 -44.7103255 90.71033 -80.55401 126.55401
## 421 23 -48.8176455 94.81765 -86.83561 132.83561
## 422 23 -52.7024453 98.70245 -92.77690 138.77690
## 423 23 -56.3973945 102.39739 -98.42784 144.42784
## 424 23 -59.9278739 105.92787 -103.82724 149.82724
## 425 23 -63.3140678 109.31407 -109.00598 155.00598
## 426 23 -66.5723412 112.57234 -113.98907 159.98907
## 427 23 -69.7161816 115.71618 -118.79716 164.79716
## 428 23 -72.7568606 118.75686 -123.44748 169.44748
## 429 23 -75.7039127 121.70391 -127.95461 173.95461
## 430 23 -78.5654883 124.56549 -132.33101 178.33101
## 431 23 -81.3486197 127.34862 -136.58744 182.58744
## 432 23 -84.0594249 130.05942 -140.73326 186.73326
accuracy(la_roche_naive,ts_la_roche_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.09821429 18.67987 13.16964 -13.36197 36.88527 1.000000
## Test set 20.00000000 28.80104 20.70000 35.18321 38.78847 1.571797
## ACF1 Theil's U
## Training set -0.1844640 NA
## Test set 0.4981374 1.12516
##mean
la_roche_mean<-meanf(ts_la_roche_train,20)
summary(la_roche_mean)
##
## Forecast method: Mean
##
## Model Information:
## $mu
## [1] 53.36283
##
## $mu.se
## [1] 2.754495
##
## $sd
## [1] 29.28068
##
## $bootstrap
## [1] FALSE
##
## $call
## meanf(y = ts_la_roche_train, h = 20)
##
## attr(,"class")
## [1] "meanf"
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set 2.762877e-15 29.15083 22.80852 -95.2591 121.4718 1.731901
## ACF1
## Training set 0.7827953
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 53.36283 15.44881 91.27685 -4.909225 111.6349
## 414 53.36283 15.44881 91.27685 -4.909225 111.6349
## 415 53.36283 15.44881 91.27685 -4.909225 111.6349
## 416 53.36283 15.44881 91.27685 -4.909225 111.6349
## 417 53.36283 15.44881 91.27685 -4.909225 111.6349
## 418 53.36283 15.44881 91.27685 -4.909225 111.6349
## 419 53.36283 15.44881 91.27685 -4.909225 111.6349
## 420 53.36283 15.44881 91.27685 -4.909225 111.6349
## 421 53.36283 15.44881 91.27685 -4.909225 111.6349
## 422 53.36283 15.44881 91.27685 -4.909225 111.6349
## 423 53.36283 15.44881 91.27685 -4.909225 111.6349
## 424 53.36283 15.44881 91.27685 -4.909225 111.6349
## 425 53.36283 15.44881 91.27685 -4.909225 111.6349
## 426 53.36283 15.44881 91.27685 -4.909225 111.6349
## 427 53.36283 15.44881 91.27685 -4.909225 111.6349
## 428 53.36283 15.44881 91.27685 -4.909225 111.6349
## 429 53.36283 15.44881 91.27685 -4.909225 111.6349
## 430 53.36283 15.44881 91.27685 -4.909225 111.6349
## 431 53.36283 15.44881 91.27685 -4.909225 111.6349
## 432 53.36283 15.44881 91.27685 -4.909225 111.6349
accuracy(la_roche_mean,ts_la_roche_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 2.762877e-15 29.15083 22.80852 -95.25910 121.47178 1.731901
## Test set -1.036283e+01 23.17085 20.68142 -50.38294 62.72762 1.570385
## ACF1 Theil's U
## Training set 0.7827953 NA
## Test set 0.4981374 1.036926
##ets
la_roche_ets<-ets(ts_la_roche_train)
checkresiduals(la_roche_ets)
##
## Ljung-Box test
##
## data: Residuals from ETS(A,N,N)
## Q* = 16.18, df = 8, p-value = 0.03987
##
## Model df: 2. Total lags used: 10
la_roche_ets_fc<-forecast(la_roche_ets,20)
autoplot(la_roche_ets_fc)
summary(la_roche_ets)
## ETS(A,N,N)
##
## Call:
## ets(y = ts_la_roche_train)
##
## Smoothing parameters:
## alpha = 0.7278
##
## Initial states:
## l = 10.8187
##
## sigma: 18.3405
##
## AIC AICc BIC
## 1195.636 1195.856 1203.818
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0.2296552 18.17745 13.04539 -13.97018 36.26489 0.9905649 0.047763
accuracy(la_roche_ets_fc,ts_la_roche_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.2296552 18.17745 13.04539 -13.97018 36.26489 0.9905649
## Test set 13.2947346 24.62214 16.21789 16.28696 30.41886 1.2314604
## ACF1 Theil's U
## Training set 0.0477630 NA
## Test set 0.4981374 0.976256
##arima
la_roche_arima<-auto.arima(ts_la_roche_train)
checkresiduals(la_roche_arima)
##
## Ljung-Box test
##
## data: Residuals from ARIMA(1,1,1)
## Q* = 6.6679, df = 8, p-value = 0.5729
##
## Model df: 2. Total lags used: 10
la_roche_arima_fc<-forecast(la_roche_arima,20)
autoplot(la_roche_arima_fc)
accuracy(la_roche_arima_fc,ts_la_roche_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.6724592 17.42081 12.75499 -13.58058522 34.78088 0.9685142
## Test set 7.5016807 22.42854 15.19963 -0.07433621 32.16189 1.1541414
## ACF1 Theil's U
## Training set 0.01387811 NA
## Test set 0.54730368 0.9381456
##tbats
la_roche_tbats<-tbats(ts_la_roche_train)
la_roche_tbats_fc<-forecast(la_roche_tbats,20)
autoplot(la_roche_tbats_fc)
accuracy(la_roche_tbats_fc,ts_la_roche_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.8680211 18.33629 13.18135 -14.121367 36.36587 1.000889
## Test set 10.4943039 23.22995 15.29886 8.395002 30.15934 1.161676
## ACF1 Theil's U
## Training set 0.2007237 NA
## Test set 0.4981374 0.9315743
##linear regression
autoplot(ts_la_roche)
la_roche<-la_roche[,-"product_content_id"]
train.la_roche<-la_roche[300:412,]
test.la_roche<-la_roche[413:432,]
la_roche_linear<-lm(sold_count~.,data=train.la_roche)
la_roche.predict<-predict(la_roche_linear,test.la_roche)
accuracy(la_roche.predict,test.la_roche$sold_count)
## ME RMSE MAE MPE MAPE
## Test set -7.349925 12.02528 9.846595 -26.78751 29.66923
plot(la_roche.predict)
##linear according to correlation
ggcorr(train.la_roche,label=TRUE)
## Warning in ggcorr(train.la_roche, label = TRUE): data in column(s) 'event_date'
## are not numeric and were ignored
la_roche_linear_corr<-lm(sold_count~basket_count+favored_count+visit_count+
category_visits+ty_visits,train.la_roche)
la_roche.predict.corr<-predict(la_roche_linear_corr,test.la_roche)
accuracy(la_roche.predict.corr,test.la_roche$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 3.409211 10.63975 7.755503 0.4453327 18.63708
plot(la_roche.predict.corr,type="l",xlab="Days",ylab="La Roche Cleanser")
####Trendyol Tayt
##Building test and train sets
ts_trend_tayt_train<-window(ts_trend_tayt,start=300,end=412)
ts_trend_tayt_test<-window(ts_trend_tayt,start=413,end=432)
##holt
trend_tayt_holt<-holt(ts_trend_tayt_train,20)
summary(trend_tayt_holt)
##
## Forecast method: Holt's method
##
## Model Information:
## Holt's method
##
## Call:
## holt(y = ts_trend_tayt_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.5918
## beta = 1e-04
##
## Initial states:
## l = 879.0846
## b = -2.0397
##
## sigma: 305.3758
##
## AIC AICc BIC
## 1833.191 1833.752 1846.828
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0.1447713 299.9222 178.4934 -Inf Inf 1.005951 0.1121222
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 654.0773 262.722477 1045.432 55.55173 1252.603
## 414 652.0393 197.266730 1106.812 -43.47532 1347.554
## 415 650.0012 139.613480 1160.389 -130.56948 1430.572
## 416 647.9632 87.435416 1208.491 -209.29006 1505.216
## 417 645.9251 39.373210 1252.477 -281.71599 1573.566
## 418 643.8871 -5.448913 1293.223 -349.18663 1636.961
## 419 641.8490 -47.634382 1331.332 -412.62486 1696.323
## 420 639.8110 -87.619889 1367.242 -472.69854 1752.321
## 421 637.7729 -125.733528 1401.279 -529.90944 1805.455
## 422 635.7349 -162.229223 1433.699 -584.64591 1856.116
## 423 633.6968 -197.308263 1464.702 -637.21580 1904.609
## 424 631.6588 -231.133420 1494.451 -687.86803 1951.186
## 425 629.6207 -263.838535 1523.080 -736.80731 1996.049
## 426 627.5827 -295.535245 1550.701 -784.20437 2039.370
## 427 625.5447 -326.317817 1577.407 -830.20337 2081.293
## 428 623.5066 -356.266713 1603.280 -874.92738 2121.941
## 429 621.4686 -385.451262 1628.388 -918.48242 2161.420
## 430 619.4305 -413.931707 1652.793 -960.96062 2199.822
## 431 617.3925 -441.760785 1676.546 -1002.44264 2237.228
## 432 615.3544 -468.984977 1699.694 -1042.99957 2273.708
autoplot(trend_tayt_holt)
accuracy(trend_tayt_holt,ts_trend_tayt_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 0.1447713 299.9222 178.4934 -Inf Inf 1.005951
## Test set -187.5658707 243.7958 229.1578 -56.63483 61.86349 1.291484
## ACF1 Theil's U
## Training set 0.1121222 NA
## Test set 0.7571179 3.977911
checkresiduals(trend_tayt_holt)
##
## Ljung-Box test
##
## data: Residuals from Holt's method
## Q* = 22.455, df = 6, p-value = 0.001001
##
## Model df: 4. Total lags used: 10
##ses
trend_tayt_ses<-ses(ts_trend_tayt_train,20)
summary(trend_tayt_ses)
##
## Forecast method: Simple exponential smoothing
##
## Model Information:
## Simple exponential smoothing
##
## Call:
## ses(y = ts_trend_tayt_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.5942
##
## Initial states:
## l = 897.8195
##
## sigma: 302.6001
##
## AIC AICc BIC
## 1829.182 1829.402 1837.364
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -3.575563 299.9103 178.9255 -Inf Inf 1.008386 0.1107847
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 657.7487 269.95099 1045.546 64.6633 1250.834
## 414 657.7487 206.66021 1108.837 -32.1316 1347.629
## 415 657.7487 151.21678 1164.281 -116.9250 1432.422
## 416 657.7487 101.27017 1214.227 -193.3117 1508.809
## 417 657.7487 55.45134 1260.046 -263.3856 1578.883
## 418 657.7487 12.87982 1302.618 -328.4931 1643.990
## 419 657.7487 -27.05027 1342.548 -389.5609 1705.058
## 420 657.7487 -64.77701 1380.274 -447.2590 1762.756
## 421 657.7487 -100.62928 1416.127 -502.0903 1817.588
## 422 657.7487 -134.86149 1450.359 -554.4440 1869.941
## 423 657.7487 -167.67524 1483.173 -604.6283 1920.126
## 424 657.7487 -199.23347 1514.731 -652.8924 1968.390
## 425 657.7487 -229.67014 1545.167 -699.4413 2014.939
## 426 657.7487 -259.09696 1574.594 -744.4457 2059.943
## 427 657.7487 -287.60823 1603.106 -788.0499 2103.547
## 428 657.7487 -315.28443 1630.782 -830.3770 2145.874
## 429 657.7487 -342.19492 1657.692 -871.5331 2187.030
## 430 657.7487 -368.39992 1683.897 -911.6102 2227.108
## 431 657.7487 -393.95219 1709.450 -950.6890 2266.186
## 432 657.7487 -418.89819 1734.396 -988.8406 2304.338
autoplot(trend_tayt_ses)+autolayer(fitted(trend_tayt_ses))
accuracy(trend_tayt_ses,ts_trend_tayt_test)
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -3.575563 299.9103 178.9255 -Inf Inf 1.008386 0.1107847
## Test set -210.598667 267.2649 249.4992 -62.99381 67.8606 1.406124 0.7664920
## Theil's U
## Training set NA
## Test set 4.398075
##naive
trend_tayt_naive<-naive(ts_trend_tayt_train,20)
summary(trend_tayt_naive)
##
## Forecast method: Naive method
##
## Model Information:
## Call: naive(y = ts_trend_tayt_train, h = 20)
##
## Residual sd: 313.4998
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -2.991071 312.1115 177.4375 -Inf Inf 1 -0.20393
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 672 272.01307 1071.987 60.27278 1283.727
## 414 672 106.33306 1237.667 -193.11293 1537.113
## 415 672 -20.79768 1364.798 -387.54262 1731.543
## 416 672 -127.97386 1471.974 -551.45444 1895.454
## 417 672 -222.39796 1566.398 -695.86364 2039.864
## 418 672 -307.76388 1651.764 -826.41955 2170.420
## 419 672 -386.26594 1730.266 -946.47809 2290.478
## 420 672 -459.33388 1803.334 -1058.22586 2402.226
## 421 672 -527.96078 1871.961 -1163.18165 2507.182
## 422 672 -592.86973 1936.870 -1262.45131 2606.451
## 423 672 -654.60656 1998.607 -1356.86966 2700.870
## 424 672 -713.59536 2057.595 -1447.08524 2791.085
## 425 672 -770.17338 2114.173 -1533.61385 2877.614
## 426 672 -824.61404 2168.614 -1616.87366 2960.874
## 427 672 -877.14271 2221.143 -1697.20933 3041.209
## 428 672 -927.94771 2271.948 -1774.90887 3118.909
## 429 672 -977.18835 2321.188 -1850.21593 3194.216
## 430 672 -1025.00082 2369.001 -1923.33878 3267.339
## 431 672 -1071.50260 2415.503 -1994.45712 3338.457
## 432 672 -1116.79592 2460.796 -2063.72729 3407.727
accuracy(trend_tayt_naive,ts_trend_tayt_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -2.991071 312.1115 177.4375 -Inf Inf 1.000000
## Test set -224.850000 278.6328 258.0500 -66.52537 70.63093 1.454315
## ACF1 Theil's U
## Training set -0.203930 NA
## Test set 0.766492 4.596106
##mean
trend_tayt_mean<-meanf(ts_trend_tayt_train,20)
summary(trend_tayt_mean)
##
## Forecast method: Mean
##
## Model Information:
## $mu
## [1] 648.0796
##
## $mu.se
## [1] 31.08111
##
## $sd
## [1] 330.3968
##
## $bootstrap
## [1] FALSE
##
## $call
## meanf(y = ts_trend_tayt_train, h = 20)
##
## attr(,"class")
## [1] "meanf"
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -1.810072e-14 328.9316 257.4354 -Inf Inf 1.450851 0.5485202
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 413 648.0796 220.2662 1075.893 -9.449438 1305.609
## 414 648.0796 220.2662 1075.893 -9.449438 1305.609
## 415 648.0796 220.2662 1075.893 -9.449438 1305.609
## 416 648.0796 220.2662 1075.893 -9.449438 1305.609
## 417 648.0796 220.2662 1075.893 -9.449438 1305.609
## 418 648.0796 220.2662 1075.893 -9.449438 1305.609
## 419 648.0796 220.2662 1075.893 -9.449438 1305.609
## 420 648.0796 220.2662 1075.893 -9.449438 1305.609
## 421 648.0796 220.2662 1075.893 -9.449438 1305.609
## 422 648.0796 220.2662 1075.893 -9.449438 1305.609
## 423 648.0796 220.2662 1075.893 -9.449438 1305.609
## 424 648.0796 220.2662 1075.893 -9.449438 1305.609
## 425 648.0796 220.2662 1075.893 -9.449438 1305.609
## 426 648.0796 220.2662 1075.893 -9.449438 1305.609
## 427 648.0796 220.2662 1075.893 -9.449438 1305.609
## 428 648.0796 220.2662 1075.893 -9.449438 1305.609
## 429 648.0796 220.2662 1075.893 -9.449438 1305.609
## 430 648.0796 220.2662 1075.893 -9.449438 1305.609
## 431 648.0796 220.2662 1075.893 -9.449438 1305.609
## 432 648.0796 220.2662 1075.893 -9.449438 1305.609
accuracy(trend_tayt_mean,ts_trend_tayt_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -1.810072e-14 328.9316 257.4354 -Inf Inf 1.450851
## Test set -2.009296e+02 259.7142 243.6978 -60.59777 65.98103 1.373429
## ACF1 Theil's U
## Training set 0.5485202 NA
## Test set 0.7664920 4.264387
##ets
trend_tayt_ets<-ets(ts_trend_tayt_train)
checkresiduals(trend_tayt_ets)
##
## Ljung-Box test
##
## data: Residuals from ETS(A,N,N)
## Q* = 22.416, df = 8, p-value = 0.0042
##
## Model df: 2. Total lags used: 10
trend_tayt_ets_fc<-forecast(trend_tayt_ets,20)
autoplot(trend_tayt_ets_fc)
summary(trend_tayt_ets)
## ETS(A,N,N)
##
## Call:
## ets(y = ts_trend_tayt_train)
##
## Smoothing parameters:
## alpha = 0.594
##
## Initial states:
## l = 897.8354
##
## sigma: 302.6001
##
## AIC AICc BIC
## 1829.182 1829.402 1837.364
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set -3.577172 299.9103 178.9278 -Inf Inf 1.008399 0.1109197
accuracy(trend_tayt_ets_fc,ts_trend_tayt_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -3.577172 299.9103 178.9278 -Inf Inf 1.008399
## Test set -210.581459 267.2513 249.4889 -62.98955 67.85725 1.406066
## ACF1 Theil's U
## Training set 0.1109197 NA
## Test set 0.7664920 4.397837
##arima
trend_tayt_arima<-auto.arima(ts_trend_tayt_train)
checkresiduals(trend_tayt_arima)
##
## Ljung-Box test
##
## data: Residuals from ARIMA(1,0,0) with non-zero mean
## Q* = 8.983, df = 8, p-value = 0.3437
##
## Model df: 2. Total lags used: 10
trend_tayt_arima_fc<-forecast(trend_tayt_arima,20)
autoplot(trend_tayt_arima_fc)
accuracy(trend_tayt_arima_fc,ts_trend_tayt_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -2.245072 274.3966 179.1991 -Inf Inf 1.009928
## Test set -206.193051 262.5491 245.1377 -61.77595 66.64837 1.381544
## ACF1 Theil's U
## Training set 0.01181821 NA
## Test set 0.76522197 4.320004
##tbats
trend_tayt_tbats<-tbats(ts_trend_tayt_train)
trend_tayt_tbats_fc<-forecast(trend_tayt_tbats,20)
autoplot(trend_tayt_tbats_fc)
accuracy(trend_tayt_tbats_fc,ts_trend_tayt_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -10.43774 273.1901 182.9158 -Inf Inf 1.030875
## Test set -234.95370 287.9357 265.7275 -69.14369 72.92566 1.497584
## ACF1 Theil's U
## Training set -0.03888996 NA
## Test set 0.76708561 4.752005
##linear regression
autoplot(ts_trend_tayt)
trend_tayt<-trend_tayt[,-"product_content_id"]
train.trend_tayt<-trend_tayt[300:412,]
test.trend_tayt<-trend_tayt[413:432,]
trend_tayt_linear<-lm(sold_count~.,data=train.trend_tayt)
trend_tayt.predict<-predict(trend_tayt_linear,test.trend_tayt)
accuracy(trend_tayt.predict,test.trend_tayt$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 98.02491 150.645 133.8769 27.7358 31.90949
plot(trend_tayt.predict)
##linear according to correlation
ggcorr(train.trend_tayt,label=TRUE)
## Warning in ggcorr(train.trend_tayt, label = TRUE): data in column(s)
## 'event_date' are not numeric and were ignored
trend_tayt_linear_corr<-lm(sold_count~basket_count+favored_count+visit_count
,train.trend_tayt)
trend_tayt.predict.corr<-predict(trend_tayt_linear_corr,test.trend_tayt)
accuracy(trend_tayt.predict.corr,test.trend_tayt$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 50.59298 114.0405 85.6518 13.50563 17.98632
plot(trend_tayt.predict.corr)
plot(trend_tayt.predict.corr,type="l",xlab="Days",ylab="Trendyol Tights")
####Oralb
##Building test and train sets
ts_oralb_train<-window(ts_oralb,start=280,end=378)
ts_oralb_test<-window(ts_oralb,start=379,end=398)
##holt
oralb_holt<-holt(ts_oralb_train,20)
summary(oralb_holt)
##
## Forecast method: Holt's method
##
## Model Information:
## Holt's method
##
## Call:
## holt(y = ts_oralb_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.5822
## beta = 1e-04
##
## Initial states:
## l = 256.6825
## b = 0.2307
##
## sigma: 100.0699
##
## AIC AICc BIC
## 1372.796 1373.441 1385.772
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set -0.4782952 98.02748 56.13349 -50.88779 75.79857 0.9688415
## ACF1
## Training set 0.02147262
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 379 250.1327 121.887920 378.3775 53.999232 446.2662
## 380 250.3586 101.954988 398.7623 23.394846 477.3224
## 381 250.5846 84.444635 416.7245 -3.504524 504.6736
## 382 250.8105 68.647856 432.9731 -27.783208 529.4042
## 383 251.0364 54.145861 447.9269 -50.081692 552.1545
## 384 251.2623 40.666822 461.8578 -70.815700 573.3403
## 385 251.4882 28.022448 474.9540 -90.273197 593.2497
## 386 251.7142 16.075933 487.3524 -108.663411 612.0917
## 387 251.9401 4.724172 499.1560 -126.144028 630.0242
## 388 252.1660 -6.112831 510.4448 -142.837390 647.1694
## 389 252.3919 -16.498619 521.2825 -158.840678 663.6245
## 390 252.6179 -26.484665 531.7204 -174.232613 679.4683
## 391 252.8438 -36.113353 541.8009 -189.078017 694.7656
## 392 253.0697 -45.420082 551.5595 -203.431026 709.5704
## 393 253.2956 -54.434772 561.0260 -217.337401 723.9286
## 394 253.5215 -63.182988 570.2261 -230.836240 737.8793
## 395 253.7475 -71.686777 579.1817 -243.961259 751.4562
## 396 253.9734 -79.965309 587.9121 -256.741776 764.6886
## 397 254.1993 -88.035377 596.4340 -269.203476 777.6021
## 398 254.4252 -95.911791 604.7623 -281.369009 790.2195
autoplot(oralb_holt)
accuracy(oralb_holt,ts_oralb_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -0.4782952 98.02748 56.13349 -50.88779 75.79857 0.9688415
## Test set -168.7289690 171.55697 168.72897 -261.76169 261.76169 2.9121943
## ACF1 Theil's U
## Training set 0.02147262 NA
## Test set 0.43052069 6.056433
checkresiduals(oralb_holt)
##
## Ljung-Box test
##
## data: Residuals from Holt's method
## Q* = 11.313, df = 6, p-value = 0.07916
##
## Model df: 4. Total lags used: 10
##ses
oralb_ses<-ses(ts_oralb_train,20)
summary(oralb_ses)
##
## Forecast method: Simple exponential smoothing
##
## Model Information:
## Simple exponential smoothing
##
## Call:
## ses(y = ts_oralb_train, h = 20)
##
## Smoothing parameters:
## alpha = 0.5891
##
## Initial states:
## l = 184.0352
##
## sigma: 98.6958
##
## AIC AICc BIC
## 1368.121 1368.373 1375.906
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 1.11778 97.69381 55.41254 -48.7249 74.04941 0.9563983 0.02667279
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 379 249.2294 122.745598 375.7132 55.789125 442.6696
## 380 249.2294 102.427274 396.0315 24.714930 473.7438
## 381 249.2294 84.597763 413.8610 -2.552955 501.0117
## 382 249.2294 68.518892 429.9399 -27.143466 525.6022
## 383 249.2294 53.758175 444.7006 -49.718033 548.1768
## 384 249.2294 40.036400 458.4224 -70.703679 569.1625
## 385 249.2294 27.160890 471.2979 -90.395071 588.8538
## 386 249.2294 14.992053 483.4667 -109.005702 607.4645
## 387 249.2294 3.424911 495.0339 -126.696118 625.1549
## 388 249.2294 -7.621838 506.0806 -143.590663 642.0494
## 389 249.2294 -18.212687 516.6715 -159.787969 658.2467
## 390 249.2294 -28.399816 526.8586 -175.367839 673.8266
## 391 249.2294 -38.226150 536.6849 -190.395920 688.8547
## 392 249.2294 -47.727508 546.1863 -204.926992 703.3858
## 393 249.2294 -56.934146 555.3929 -219.007330 717.4661
## 394 249.2294 -65.871899 564.3307 -232.676443 731.1352
## 395 249.2294 -74.563034 573.0218 -245.968387 744.4272
## 396 249.2294 -83.026904 581.4857 -258.912760 757.3715
## 397 249.2294 -91.280458 589.7392 -271.535481 769.9943
## 398 249.2294 -99.338635 597.7974 -283.859400 782.3182
autoplot(oralb_ses)+autolayer(fitted(oralb_ses))
accuracy(oralb_ses,ts_oralb_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 1.11778 97.69381 55.41254 -48.7249 74.04941 0.9563983
## Test set -165.67939 168.42537 165.67939 -256.8314 256.83142 2.8595597
## ACF1 Theil's U
## Training set 0.02667279 NA
## Test set 0.40785154 5.933022
##naive
oralb_naive<-naive(ts_oralb_train,20)
summary(oralb_naive)
##
## Forecast method: Naive method
##
## Model Information:
## Call: naive(y = ts_oralb_train, h = 20)
##
## Residual sd: 105.3442
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 1.22449 104.8125 57.93878 -28.63994 59.04047 1 -0.3212452
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 379 225 90.677420 359.3226 19.57134 430.4287
## 380 225 35.039186 414.9608 -65.52000 515.5200
## 381 225 -7.653533 457.6535 -130.81287 580.8129
## 382 225 -43.645160 493.6452 -185.85732 635.8573
## 383 225 -75.354420 525.3544 -234.35245 684.3524
## 384 225 -104.021782 554.0218 -278.19539 728.1954
## 385 225 -130.384142 580.3841 -318.51314 768.5131
## 386 225 -154.921629 604.9216 -356.03999 806.0400
## 387 225 -177.967740 627.9677 -391.28598 841.2860
## 388 225 -199.765294 649.7653 -424.62246 874.6225
## 389 225 -220.497599 670.4976 -456.32978 906.3298
## 390 225 -240.307067 690.3071 -486.62575 936.6257
## 391 225 -259.306950 709.3069 -515.68356 965.6836
## 392 225 -277.589074 727.5891 -543.64366 993.6437
## 393 225 -295.229116 745.2291 -570.62177 1020.6218
## 394 225 -312.290320 762.2903 -596.71464 1046.7146
## 395 225 -328.826186 778.8262 -622.00406 1072.0041
## 396 225 -344.882443 794.8824 -646.55999 1096.5600
## 397 225 -360.498552 810.4986 -670.44276 1120.4428
## 398 225 -375.708840 825.7088 -693.70489 1143.7049
accuracy(oralb_naive,ts_oralb_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 1.22449 104.8125 57.93878 -28.63994 59.04047 1.00000
## Test set -141.45000 144.6567 141.45000 -222.14126 222.14126 2.44137
## ACF1 Theil's U
## Training set -0.3212452 NA
## Test set 0.4078515 5.159023
##mean
oralb_mean<-meanf(ts_oralb_train,20)
summary(oralb_mean)
##
## Forecast method: Mean
##
## Model Information:
## $mu
## [1] 146.4747
##
## $mu.se
## [1] 11.90618
##
## $sd
## [1] 118.465
##
## $bootstrap
## [1] FALSE
##
## $call
## meanf(y = ts_oralb_train, h = 20)
##
## attr(,"class")
## [1] "meanf"
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE
## Training set -1.235053e-14 117.8652 81.38659 -262.7059 288.7156 1.4047
## ACF1
## Training set 0.6057377
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 379 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 380 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 381 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 382 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 383 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 384 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 385 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 386 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 387 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 388 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 389 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 390 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 391 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 392 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 393 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 394 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 395 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 396 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 397 146.4747 -7.144789 300.0943 -89.79961 382.7491
## 398 146.4747 -7.144789 300.0943 -89.79961 382.7491
accuracy(oralb_mean,ts_oralb_test)
## ME RMSE MAE MPE MAPE MASE
## Training set -1.235053e-14 117.86522 81.38659 -262.7059 288.7156 1.404700
## Test set -6.292475e+01 69.83532 62.92475 -109.7136 109.7136 1.086056
## ACF1 Theil's U
## Training set 0.6057377 NA
## Test set 0.4078515 2.725752
##ets
oralb_ets<-ets(ts_oralb_train)
checkresiduals(oralb_ets)
##
## Ljung-Box test
##
## data: Residuals from ETS(A,N,N)
## Q* = 12.327, df = 8, p-value = 0.1372
##
## Model df: 2. Total lags used: 10
oralb_ets_fc<-forecast(oralb_ets,20)
autoplot(oralb_ets_fc)
summary(oralb_ets)
## ETS(A,N,N)
##
## Call:
## ets(y = ts_oralb_train)
##
## Smoothing parameters:
## alpha = 0.5892
##
## Initial states:
## l = 183.8392
##
## sigma: 98.6958
##
## AIC AICc BIC
## 1368.121 1368.373 1375.906
##
## Training set error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 1.12102 97.69381 55.41076 -48.72023 74.04525 0.9563674 0.02667636
accuracy(oralb_ets_fc,ts_oralb_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 1.12102 97.69381 55.41076 -48.72023 74.04525 0.9563674
## Test set -165.67658 168.42261 165.67658 -256.82740 256.82740 2.8595112
## ACF1 Theil's U
## Training set 0.02667636 NA
## Test set 0.40785154 5.932932
##arima
oralb_arima<-auto.arima(ts_oralb_train)
checkresiduals(oralb_arima)
##
## Ljung-Box test
##
## data: Residuals from ARIMA(2,1,1)
## Q* = 6.7455, df = 7, p-value = 0.4558
##
## Model df: 3. Total lags used: 10
oralb_arima_fc<-forecast(oralb_arima,20)
autoplot(oralb_arima_fc)
accuracy(oralb_arima_fc,ts_oralb_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 7.112649 92.90530 51.96090 -75.04919 97.53398 0.8968243
## Test set -94.160322 98.32862 94.16032 -152.07644 152.07644 1.6251693
## ACF1 Theil's U
## Training set 0.01587862 NA
## Test set 0.22154226 3.513086
##tbats
oralb_tbats<-tbats(ts_oralb_train)
oralb_tbats_fc<-forecast(oralb_tbats,20)
autoplot(oralb_tbats_fc)
accuracy(oralb_tbats_fc,ts_oralb_test)
## ME RMSE MAE MPE MAPE MASE
## Training set 6.939308 97.43357 52.82237 -30.20269 59.03135 0.9116929
## Test set -154.875983 157.81007 154.87598 -241.36377 241.36377 2.6730973
## ACF1 Theil's U
## Training set -0.01075659 NA
## Test set 0.40785154 5.587261
##linear regression
autoplot(ts_oralb)
oralb<-oralb[,-"product_content_id"]
train.oralb<-oralb[280:378,]
test.oralb<-oralb[379:388,]
oralb_linear<-lm(sold_count~.,data=train.oralb)
oralb.predict<-predict(oralb_linear,test.oralb)
accuracy(oralb.predict,test.oralb$sold_count)
## ME RMSE MAE MPE MAPE
## Test set -64.89802 67.68503 64.89802 -74.82121 74.82121
plot(oralb.predict)
##linear according to correlation
ggcorr(train.oralb,label=TRUE)
## Warning in ggcorr(train.oralb, label = TRUE): data in column(s) 'event_date' are
## not numeric and were ignored
oralb_linear_corr<-lm(sold_count~basket_count+favored_count+visit_count+
category_visits,train.oralb)
oralb.predict.corr<-predict(oralb_linear_corr,test.oralb)
accuracy(oralb.predict.corr,test.oralb$sold_count)
## ME RMSE MAE MPE MAPE
## Test set -47.97082 52.73447 47.97082 -57.36166 57.36166
plot(oralb.predict.corr,type="l",xlab="Days",ylab="Oral-B Toothbrush")
####Koton Mont
##Building test and train sets
ts_koton_train<-window(ts_koton,start=300,end=395)
ts_koton_test<-window(ts_koton,start=396,end=415)
##naive
koton_naive<-naive(ts_koton_train,20)
summary(koton_naive)
##
## Forecast method: Naive method
##
## Model Information:
## Call: naive(y = ts_koton_train, h = 20)
##
## Residual sd: 0.2526
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0 0.2513123 0.06315789 -Inf Inf 1 -0.5
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 396 0 -0.3220697 0.3220697 -0.4925631 0.4925631
## 397 0 -0.4554754 0.4554754 -0.6965895 0.6965895
## 398 0 -0.5578411 0.5578411 -0.8531444 0.8531444
## 399 0 -0.6441395 0.6441395 -0.9851263 0.9851263
## 400 0 -0.7201698 0.7201698 -1.1014047 1.1014047
## 401 0 -0.7889065 0.7889065 -1.2065284 1.2065284
## 402 0 -0.8521164 0.8521164 -1.3031996 1.3031996
## 403 0 -0.9109508 0.9109508 -1.3931790 1.3931790
## 404 0 -0.9662092 0.9662092 -1.4776894 1.4776894
## 405 0 -1.0184739 1.0184739 -1.5576214 1.5576214
## 406 0 -1.0681844 1.0681844 -1.6336471 1.6336471
## 407 0 -1.1156823 1.1156823 -1.7062888 1.7062888
## 408 0 -1.1612389 1.1612389 -1.7759617 1.7759617
## 409 0 -1.2050746 1.2050746 -1.8430025 1.8430025
## 410 0 -1.2473707 1.2473707 -1.9076889 1.9076889
## 411 0 -1.2882789 1.2882789 -1.9702526 1.9702526
## 412 0 -1.3279275 1.3279275 -2.0308899 2.0308899
## 413 0 -1.3664261 1.3664261 -2.0897684 2.0897684
## 414 0 -1.4038694 1.4038694 -2.1470330 2.1470330
## 415 0 -1.4403396 1.4403396 -2.2028094 2.2028094
accuracy(koton_naive,ts_koton_test)
## ME RMSE MAE MPE MAPE MASE ACF1 Theil's U
## Training set 0 0.2513123 0.06315789 -Inf Inf 1 -0.5 NA
## Test set 0 0.0000000 0.00000000 NaN NaN 0 NaN NaN
####Trendyol bikini
##Building test and train sets
ts_trend_bikini_train<-window(ts_trend_bikini,start=300,end=395)
ts_trend_bikini_test<-window(ts_trend_bikini,start=396,end=415)
##holt
trend_bikini_holt<-holt(ts_trend_bikini_train,20)
accuracy(trend_bikini_holt,ts_trend_bikini_test)
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0.01615299 0.5805658 0.4359376 NaN Inf 0.900306 -0.001865744
## Test set 0.07684537 0.8235117 0.6864552 -Inf Inf 1.417679 -0.048783309
## Theil's U
## Training set NA
## Test set 0
##mean
trend_bikini_mean<-meanf(ts_trend_bikini_train,20)
summary(trend_bikini_mean)
##
## Forecast method: Mean
##
## Model Information:
## $mu
## [1] 0.3333333
##
## $mu.se
## [1] 0.06045635
##
## $sd
## [1] 0.5923489
##
## $bootstrap
## [1] FALSE
##
## $call
## meanf(y = ts_trend_bikini_train, h = 20)
##
## attr(,"class")
## [1] "meanf"
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 3.351145e-17 0.5892557 0.4861111 -Inf Inf 1.003925 0.06666667
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 396 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 397 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 398 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 399 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 400 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 401 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 402 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 403 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 404 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 405 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 406 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 407 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 408 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 409 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 410 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 411 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 412 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 413 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 414 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
## 415 0.3333333 -0.4350798 1.101746 -0.8487368 1.515403
accuracy(trend_bikini_mean,ts_trend_bikini_test)
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 3.351145e-17 0.5892557 0.4861111 -Inf Inf 1.003925 0.06666667
## Test set 4.166667e-01 0.9279607 0.7166667 -Inf Inf 1.480072 -0.04090909
## Theil's U
## Training set NA
## Test set 0
##seasonal naive
trend_bikini_snaive<-snaive(ts_trend_bikini_train,20)
summary(trend_bikini_snaive)
##
## Forecast method: Seasonal naive method
##
## Model Information:
## Call: snaive(y = ts_trend_bikini_train, h = 20)
##
## Residual sd: 0.8121
##
## Error measures:
## ME RMSE MAE MPE MAPE MASE ACF1
## Training set 0 0.8078562 0.4842105 -Inf Inf 1 -0.4677419
##
## Forecasts:
## Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
## 396 0 -1.035309 1.035309 -1.583369 1.583369
## 397 0 -1.464148 1.464148 -2.239222 2.239222
## 398 0 -1.793208 1.793208 -2.742476 2.742476
## 399 0 -2.070619 2.070619 -3.166738 3.166738
## 400 0 -2.315022 2.315022 -3.540521 3.540521
## 401 0 -2.535980 2.535980 -3.878446 3.878446
## 402 0 -2.739171 2.739171 -4.189201 4.189201
## 403 0 -2.928297 2.928297 -4.478444 4.478444
## 404 0 -3.105928 3.105928 -4.750107 4.750107
## 405 0 -3.273936 3.273936 -5.007052 5.007052
## 406 0 -3.433733 3.433733 -5.251441 5.251441
## 407 0 -3.586417 3.586417 -5.484951 5.484951
## 408 0 -3.732861 3.732861 -5.708918 5.708918
## 409 0 -3.873773 3.873773 -5.924424 5.924424
## 410 0 -4.009736 4.009736 -6.132362 6.132362
## 411 0 -4.141237 4.141237 -6.333476 6.333476
## 412 0 -4.268690 4.268690 -6.528398 6.528398
## 413 0 -4.392445 4.392445 -6.717666 6.717666
## 414 0 -4.512809 4.512809 -6.901745 6.901745
## 415 0 -4.630044 4.630044 -7.081041 7.081041
accuracy(trend_bikini_snaive,ts_trend_bikini_test)
## ME RMSE MAE MPE MAPE MASE ACF1 Theil's U
## Training set 0.00 0.8078562 0.4842105 -Inf Inf 1.000000 -0.46774194 NA
## Test set 0.75 1.1180340 0.7500000 100 100 1.548913 -0.04090909 0
autoplot(trend_bikini_snaive)
##linear regression
trend_bikini<-trend_bikini[,-"product_content_id"]
train.trend_bikini<-trend_bikini[300:395,]
test.trend_bikini<-trend_bikini[396:415,]
trend_bikini_linear<-lm(sold_count~.,data=train.trend_bikini)
trend_bikini.predict<-predict(trend_bikini_linear,test.trend_bikini)
accuracy(trend_bikini.predict,test.trend_bikini$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 0.1666973 0.6481107 0.375193 NaN Inf
plot(trend_bikini.predict)
##linear according to correlation
ggcorr(train.trend_bikini,label=TRUE)
## Warning in ggcorr(train.trend_bikini, label = TRUE): data in column(s)
## 'event_date' are not numeric and were ignored
trend_bikini_linear_corr<-lm(sold_count~price,data=train.trend_bikini)
trend_bikini.predict.corr<-predict(trend_bikini_linear_corr,test.trend_bikini)
accuracy(trend_bikini.predict.corr,test.trend_bikini$sold_count)
## ME RMSE MAE MPE MAPE
## Test set 0.009608758 0.4770074 0.2877316 -Inf Inf